Fault diagnosis method for rolling bearings based on improved AlexNet model
A rolling bearing fault diagnosis method based on an improved Alexnet model was proposed to address the issues of difficult feature extraction,poor diagnostic accuracy,and complex model structure in traditional methods.Firstly,the vibration signals of each fault of the rolling bearing were converted into a two-dimensional feature map sample set rich in time-frequency information,and divided into a training set and a validation set in a certain proportion.Then,improvements were made to the slow training speed and low accuracy of the Alexnet model.The ImageNet image dataset was used to train the improved model and save the knowledge obtained during the training process.Finally,the saved training information was transferred and applied to the improved model for diagnosing bearing faults in the dataset.The optimization effect of the improved model was demonstrated through training and validation on some cifar-10 image datasets before and after improvement.Compared with common network models for bearing 10 category fault diagnosis,the proposed method had better diagnostic efficiency and accuracy.
rolling bearingsfault diagnosisimproved AlexNet modeltransfer learningtime frequency characteristic map
NETL
NSTL
万方数据
